Chapter 11(1e), Ch. 10 (2/3e)
Hypothesis Testing Using the Chi Square (χ2) Distribution
Outline:
The basic logic of Chi Square. The terminology used with bivariate tables. The computation of Chi Square with an
example problem using the Five Step Model
Basic Logic Chi Square is a test of significance based on
bivariate tables.
We are looking for significant differences
between the actual cell frequencies in a table (fo)
and those that would be expected by random
chance (fe).
The data are often presented in a table format. If
starting with raw data on two variables, a
bivariate table must be created first.
Bivariate Tables:
Must have a title. Cells are intersections of columns and
rows. Subtotals are called marginals. N is reported at the intersection of row and
column marginals.
Tables (cont.)
Columns are scores of the independent variable. There will be as many columns as there are
scores on the independent variable.
Rows are scores of the dependent variable. There will be as many rows as there are scores
on the dependent variable.
Tables (cont.)
There will be as many cells as there are scores on the two variables combined.
Each cell reports the number of times each combination of scores occurred.
What your table should look like:
Title
Rows Columns Total
Row 1 cell a cell b Row Marginal 1
Row 2 cell c cell d Row Marginal 2
Total Column Marginal 1
Column Marginal 2
N
The Chi Square Distribution The chi square distribution is asymmetric and its
values are always positive (Appendix C). Degrees of freedom are based on the table and are calculated as (rows-1)X(columns-1).
Example: 1e #11.2 (#10.2 in 2/3e) Question: Are the homicide rate and volume of gun
sales related for a sample of 25 cities?
The bivariate table showing the relationship between homicide rate (columns) and gun sales (rows). This 2x2 table has 4 cells.
GUN SALES Low High Totals
High 8 5 13
Low 4 8 12
Totals 12 13 N = 25
HOMICIDE RATE
Solution Using 5-Step MethodStep 1 Make Assumptions and Meet Test
Requirements Independent random samples Level of measurement is nominal Note that no assumption is made about the
shape of the sampling distribution. When the distribution is normal, a parametric test (Z- or t-test, ANOVA) can be used.
The chi square test is non-parametric. It can be used when normality is not assumed.
Step 2 State the Null and Alternate Hypothesis
H0: The variables are independent You can also say: H0: fo = fe
H1: The variables are dependent Or: H1: fo ≠ fe
Step 3 Select the Sampling Distribution and Establish the Critical Region Because normality is not assumed and our
data are in tabular form, our Sampling Distribution = χ2
Alpha = .05 df = (r-1)(c-1) = 1 χ2 (critical) = 3.841
Step 4 Calculate the Test Statistic Formula:
χ2 (obtained) =
Method:
1. Find expected frequencies for each cell.
2. Complete computational table to find χ2 (obtained)
e
eo
f
ff 2)(
1. Find expected frequencies for each cell. To find fe =
Multiply column and row marginals for each cell and divide by N.
(13*12)/25 = 156/25 = 6.24 (13*13)/25 = 169/25 = 6.76 (12*12)/25 = 144/25 = 5.76 (12*13)/25 = 156/25 = 6.24
N
marginalcolumn marginal row
Observed and Expected Frequencies for each cell (Note that totals are unchanged):
GUN SALES Low High Total
High fo = 8
fe = 6.24fo = 5
fe = 6.7613
Lowfo = 4
fe = 5.76fo = 8
fe = 6.2412
Total 12 13 N = 25
HOMICIDE RATE
2. Complete Computational Table A table like this will help organize the computations: (a) Add values for fo and fe for each cell to table.
fo fe fo - fe (fo - fe)2 (fo - fe)2 /fe
8 6.24
5 6.76
4 5.76
8 6.24
Total 25 25
Computational Table (cont.)
(b) Subtract each fe from each fo. The total of this column must be zero.
fo fe fo - fe (fo - fe)2 (fo - fe)2 /fe
8 6.24 1.76
5 6.76 -1.76
4 5.76 -1.76
8 6.24 1.76
Total 25 25 0
Computational Table (cont.) (c) Square each of these values
fo fe fo - fe (fo - fe)2 (fo - fe)2 /fe
8 6.24 1.76 3.10
5 6.76 -1.76 3.10
4 5.76 -1.76 3.10
8 6.24 1.76 3.10
Total 25 25 0
Computational Table (cont.) (d) Divide each of the squared values by the fe for that cell. (e) The sum of this column is chi square.
χ2 (obtained) = 2.00
fo fe fo - fe (fo - fe)2 (fo - fe)2 /fe
8 6.24 1.76 3.10 .50
5 6.76 -1.76 3.10 .46
4 5.76 -1.76 3.10 .54
8 6.24 1.76 3.10 .50
Total 25 25 0 χ2 = 2.00
Step 5 Make a Decision and Interpret the Results of the Test χ2 (critical) = 3.841 χ2 (obtained) = 2.00 The test statistic is not in the Critical Region.
Fail to reject the H0. There is no significant relationship between
homicide rate and gun sales.
Interpreting Chi Square
The chi square test tells us only if the variables are independent or not.
It does not tell us the pattern or nature of the relationship.
To investigate the pattern, compute % within each column and compare across the columns.
Interpreting Chi Square (cont.) Cities low on homicide rate were high in gun sales and cities
high in homicide rate were low in gun sales. As homicide rates increase, gun sales decrease. This
relationship is not significant but does have a clear pattern.
GUN SALES Low High Total
High 8 (66.7%) 5 (38.5%) 13
Low 4 (33.3%) 8 (61.5%) 12
Total 12 (100%) 13 (100%) N = 25
HOMICIDE RATE
The Limits of Chi Square
Like all tests of hypothesis, chi square is sensitive to sample size. As N increases, obtained chi square increases. With large samples, trivial relationships may be
significant. To correct for this, when N>1000, set your alpha = .01.
Remember: significance is not the same thing as importance.
Yates’ Correction for Continuity The chi square statistic is sensitive to small cell
sizes. Whenever any of your cell sizes are <5, a slightly modified formula using Yates’ correction for continuity to calculate chi square should be calculated.
Modified Formula (Note: .5 is deducted from the absolute value of fo – fe before squaring)
(obtained) =
e
eo
f
ff 2)5.0(2c
Using SPSS to Calculate Chi Square1. Open data set in SPSS.
2. Go to Analyze>Descriptive Statistics>Crosstabs.3. Move your dependent variable into the Rows box,
and your independent variable into the Columns box.
4. Click Statistics and check box for Chi square.5. Click Cells and select Column in percentages box.6. Click Continue and OK.
Note: The “rule of thumb” for analyzing your % data is “Percentage Down, Compare Across” When analyzing your %, always compare the categories of your dependent (Row) variable across the columns of your independent (Column) variable.
Practice Question:
Try 1e #11.3 (#10.3 in 2/3e) parts a and b.
Use the five step model to calculate a full solution to this question.